I decided the other day that it’d be nice to evaluate, in a sort of scientific manner, the different ISPs that operate in Liberia. People always seem to be asking for internet recommendations, and so far all I know is my own experience with my ISP. Not that that’s even close to scientific.
So a bit about methodology. Ideally I’d love to test one ISP for a week straight running quantitative tests like download speeds and measure up time. I’d like to look at how the ISP works at 2am vs, 2pm, as most providers seem to get really slow in the middle of the day. Because all ISPs in Liberia are Wireless ISPs (WISPs) I’d also love to test how a provider works in Congo Town verses Mamba Point, since location will effect signal strength and user density. Sadly, I don’t have the resources or time to do such an intensive test of each ISP. Mostly I’m borrowing friend’s USB modems when and where I can to conduct these tests. Tests were designed to be quick so I could return the borrowed hardware in a reasonable amount of time.
With this in mind, please realize that this won’t be a super accurate view of how the ISPs perform on average. It’ll be more like a one time snapshot than an evaluation of how they perform all month long. And as we know from the recent Cellcom lightning outage, “anomalies” can bring down a whole system for an extended period of time.
So I decided to have 2 tests. I wanted to do another test, but that fell through, more on that later.
- Large Download Test – A 1.9mb image file is downloaded and timed. The idea here is to see how quickly something can be downloaded that’s large enough to give the network time to get up to speed and stabilize. When a data stream starts up the network has to figure out the best way to route it, also TCP flow control and error checking also work best when a steady stream of data is present. The main goal of this test is to see what the bandwidth is, or how much data can be pushed through over time. The picture downloaded was the sunset from my trip to Harper.
- Ping Test – Ping is a super simple network diagnostic tool used to check that a message can be sent to, and a reply received from, a device on a network. On my laptop I have ping setup to run all the time so I know when the internet drops out. Ping also tells you how long it took for the reply to be received. This allows us to test latency. Latency is the measure of delay in the network. From the moment you press enter, to the time the first bytes of a web page are received by your computer, that amount of time is measured as latency. Because all of Liberia is connected to the world through VSATs, the latency will never be below 500ms (milliseconds) or half a second. This is why you hear a delay when talking to your friends abroad. The VSAT signal has to go from Liberia, to space, and then back down to Earth, probably Lebanon, Israel, or the UK. Which means the signal travels 70,000km just to reach the internet. The speed of light is roughly 300,000km/sec, so that means that just to reach the internet there’s a delay of .25 seconds, or 250ms. Then the reply has to come back from the internet, which is another .25 seconds. So you get a total minimum delay of .5 seconds or 500ms. For way more info on this VSAT lag see this article.
I wanted to test how long it took to load a medium sized web page, around 100kb with images, CSS, java script and all. But the page that I thought was static that i used for my test somehow changed sizes between tests. So I had to throw that data out, which sucks, but that’s what you get in this age of dynamic content.
Also note that I use thousands of bits per second (kbps), when measure network speeds. This is different than thousands of bytes per second. Network speeds are usually measured in bits and files are usually measured in bytes. One bit is one “off” or “on” signal, one “1”, or “0” where a byte is 8 bits. For example on a computer the letter ‘A’ is represented by these 8 bits “01000001.” Those 8 bits make up one byte.
So if you have a 128kbps internet connection and you want to download a file that is one megabyte, it’ll take (1,000,000 * 8 ) / 128,000 = 62.5 seconds to download. Not 1,000,000/128,000 = 7.8 seconds. So just keep that in mind. Also that math is over simplified because it doesn’t account for TCP/IP overhead and rounding in the measure of a megabyte, but it should get the point across about a byte vs a bit.
At my apartment we pay $150 a month for Comium’s 256kbps connection. Back in February Comium had a two for one deal that if you paid for 3 months of service you’d get twice the speed. So we should really be paying $150 a month for 128kbps or $300 a month for 256kbps. But till July we’ve got 256kbps for only$37.50 per roommate. In this comparison Comium is used as a baseline. Unlike the other ISPs here which use cellular technologies, Comium uses a technology that’s closer to super long range Wi-Fi or Wi-Max. You also have to pay $350 for a special modem that plugs in the wall and makes Comium a stationary source of internet. All the other ISPs in this comparison use cellular technologies like GPRS or CDMA which allow you to take your internet wherever you go, and their modems are quite less costly. Also Comium is only available in Monrovia, where Lonestar and Cellcom will work wherever they have a tower, which is most of the population centers in the country.
I use this at home and it’s been pretty darn fast for Liberia. At night, when the rest of Monrovia sleeps, and isn’t using the internet, I can almost stream Youtube videos. We have had some issues with the internet randomly dropping out. Usually it’s fixed by resetting the router, which makes me wonder if the problem is Comium or my router.
These numbers were tested between 6:44PM and 6:55PM from my secret lair in the heart of Mamba Point . Here are the numbers
|Large Download Test|
|Average Time:||1:22(82 seconds)|
|Ping messages dropped by the network||6 out of 620 / 0%|
Cellcom offers unlimited use of their cellular EDGE for $60 a month. You can use any EDGE device to access their service. This could include your EDGE cellphone or the USB EDGE modem Cellcom sells for $50. To get on the Cellcom EDGE data network set the following on your device:
Access Point Name (APN): web.cellcomnet.net Dial Number *99#
Generally people I’ve talked to have had good things to say about Cellcom. They say it’s reliable and fast enough. Most people I talk to do say that Libtelco is faster. This is often attributed to the fact that Libtelco has far less customers and thus their internet connection is split between less people.
Cellcom’s data was collected between 6:05pm and 7:01pm on a roof top bar overlooking mamba point.
|Large Download Test|
|Average Time:||8:17(497 seconds)|
|Ping messages dropped by the network||272 out of 2006 / 13%|
Lonestar also lets you have unlimited access of their cellular data network for $60 a month. You can also use it at the rate of $1 an hour. I use the later option all the time when I want to check email and I don’t have any other means. I can check my email in about a minute, and at a $1 that’s way cheaper than paying $60 a month. Lonestar will also sell you a USB modem for $50, or you can use your favorite mobile device instead. My ThinkPad laptop has a built in GSM modem, like a cellphone, the SIM card sits annoyingly right behind the battery, so I use Lonestar on my laptop, again because I don’t want to commit to $60 a month.
Another point of interest is that the Lonestar USB modem sold at the Lonestar office is just a Huawei EG162G modem with some Lonestar branding. Huawei E series GSM modems are supported by FrontlineSMS. So if you want to start using SMS in cool ways with your organization, just go to lonestar, get one of their modems and download FrontlineSMS.
To get Lonestar to work on your mobile device use these settings:
Access Point Name (APN): internetlcc
Lonestar was tested from my base of covert operations in Mamba Point between
|Large Download Test|
|Average Time:||4:37(277 seconds)|
|Ping messages dropped by the network||32 out of 593 / 5%|
I’m hoping to test a friends Libtelco USB modem soon. I also want to apologize for screwing up the web page test. I think having those results would be really helpful. And finally, I want to reiterate that these results are just a snap shot of the performance of each ISP, at a specific time and place. People have reported far different results at different times and places.
Special thanks to Nate and Elie for them letting me borrow their internet devices.